Upsampling PCM or DSD in 2022


The purpose of this post is to ask the collective about the best options for upsampling today, and whether it’s worth doing. I stumbled into this topic after recently considering Paul McGowan’s take on DSD, and after reflecting on the upsampling in my home theater system.

Paul believes that DSD is world’s better than PCM. That caught my attention, because, until now, I have been operating under John Darko’s view that high res (i.e. 24-bit and above) is nice but not worth chasing--CD quality is good enough. But audio does seem analogous video. And 1080p isn't good enough for my video streams. So, I now want to give DSD and upsampling a shot.

In my home theater system, I use an Nvidia Shield TV streamer for its AI upsampling, driven by its graphics processor. Plenty of video content is still being released or only available at 1080p. Although upsampled 1080p isn’t as good as 4k, it’s better than basic 1080p. That upsampling makes a big difference for me. I strongly dislike watching 1080p content on my 4k TV. As far as I’m aware, the Nvidia Shield streamer offers the best video upsampling on the market, and it only costs >$200. It occurred to me that I might get similar gratification by upsampling audio too.

Upsampling can be performed at the DAC, streamer, server, or somewhere in between. Here are the major options I’ve considered so far:

  1. PS Audio’s Direct Stream DAC costs 6k. As an FPGA unit, it has lots of extra power that this manufacturer directs towards upsampling, and upsampling PCM to DSD is a major selling point for this device. Unfortunately, you have to get to the Direct Stream in the lineup to experience that feature.
  2. Chord’s Hugo M Scaler costs 5k. Although Chord builds FPGA DACs too, Chord sells a separate component for upscaling. In the audio chain, the M Scaler sits between a streamer and DAC. Because Chord separates out upsampling functionality into its own component, Chord’s solution is likely more expensive than PS Audio’s.
  3. HQ Player software costs >$300. HQ Player is a software service that can be installed on a server, or on a computer that sits between the server and streamer. Besides being affordable, you can pair HQ Player with Roon. The author of AudioBacon reports HQ Player introducing up to 30 seconds of lag to an audio stream when set to the most demanding upsampling algorithm even when used with a powerful Windows computer. But another commenter mentioned that his M1 Apple device introduced zero lag when running HQ Player.

Because the retail price of the components in my stereo system came out to about 5k (when new), HQ Player is where I’m looking for now. Please comment if:

  • You know about some other upsampling options I ought to consider;
  • You have opinions about the value of high res audio or upsampling; or
  • If you have anything you’d like to add to the conversation.
128x128classdstreamer

I watched a Mark Levinson YTvideo yesterday. Seems he can developed a whole new take on digital decoding called C Wave

Audio legend Mark Levinson, founder of Daniel Hertz, and his team have developed disruptive audio technology that delivers a new level of natural sound and emotion in music reproduction from all formats including streaming.

 

With this new technology, Daniel Hertz builds Maria - a single chassis complete amplifier solution with uncompromised performance, elegant design, build quality, compact size, light weight, low heat and fully programmable signal flow, priced within reach of more people. Daniel Hertz speakers offer a choice of sizes and price points for listeners with different needs.

 

THE SECRET SAUCE: Daniel Hertz patented C Wave technology enables pure analog sound and emotion from digital recordings including streaming. See the video. HERE WILL BE VIDEO

 

“The Daniel Hertz Maria is the best amplifier I know how to build, without regard for price

Reviving an old thread, in case anyone is interested...

I was 100% NOS camp, either using Roon or in my R2RDAC, Audio-GD R7 HEmk3. No matter what I did, oversampling brought noise and made the sound less organic. I am running ROCK from a NUC and streaming through a Melco N1 with an external Mutec clock. Then I did two things:

1) I isolated the NUC from the SOtM switch using fiber, with Finisar SFPs and a linear power supply.

2) I purchased a used Telegartner switch and the Initio power supply and placed it just before the streamer.

Now the upsampling in PCM is doing something positive, bringing some detail and nuances that were not present before. I have not tested with DSD yet.

@rms456 There is a detailed thread just on HQP filters on community.Roonlabs.com: Which HQPlayer Filter Are You Using

There is quite a bit of info in the HQP manual as well. There is also a detailed, dedicated thread on audiophilestyle.com.

25m Ethernet run seems long. Have you tried fiber?

Most users have HQPlayer desktop on the Mac Mini M1 / M2 (which connects via Ethernet to the rest of your network) and then use a separate NAA (like the Sonore opticalRendu or SoTM SMS-200Ultra Neo) that connects to your DAC.

 

@reg19 how did you determine those settings? I asked Jussi many times years ago, never received an answer as to which filter or mod did what. Followed discussions at Audiophile Style, but no real clues.

Although I do have a computer that's capable to upsample to DSD512 (probably higher), HQPLAYER doesn't DLNA, the current standard is PC +  Audirvana  -> 25m Ethernet ->  EtherRegen-> Lumin U1. Tried NAA on mac mini and it sounded eeeuuuwww. 

@rms456 if you’re only going to DSD128, you’re not getting the HQP benefit. Perhaps, you need more computer power?

I use a M1 Mac Studio. On HQP5 (much better than HQP4), I find that going to DSD512 using AMSDM7EC 512+fs / Sinc mGA gives a very smooth analog sound with tremendous layering / soundstage. Or, sinc-long with PCM768 / LNS15. That’s it. Two settings.

@reg19 , yes finding that one setting was the undoing, could never find the right combo. Upsampled to DSD128, try this and that on headphones,  took a few weeks of tweaking time. 

Spat the dummy one day just had enough.

@rms456 I think that one narrows down to perhaps 1 PCM setting and 1 DSD setting for your equipment / room with HQP and not on a per album basis.

@reg19 , therein lies the issue with HQPlayer, there are many choices of modulators and sampling rates to choose from. I found the process difficult to narrow down for all different types of music, there was no ‘one’ setting that 

a) worked for the DAC

b) different modulators worked against each other for say jazz, 70’s rock, or vocals. 
 

The process was such that tuning HQPlayer was needed per album. That was the time to stop tinkering and listening instead. For years now, I just let the DAC convert PCM/DSD the sample rate of the album, done. The Playback Designs MPD-3 DAC I own internally converts anything to DSD, I liked the sound of that DAC, it now has a fault, so it’s a dust collector. 
The current DAC, Accuphase DC-950 converts anything, it doesn’t upsample to DSD, whatever they use works really well, very happy with it for many years now.

@hgeifman HQP does not do the same thing as the PS Audio new DAC does. It does not just upsample to DSD.

It gives you a vast choice of modulators and filters to suit your room and equipment. That is its defining factor.

@hgeifman

I started this thread over a year ago and have mentally made note of other audio upsampling options. Besides offerings from Chord and PS Audio, Lumin streamers support upscaling to DSD, Emm labs DAC upsamples to DSD. I forget what others upsample off hand.

I like watching Steve Huff’s reviews, and in this review of the Lumin U2, Huff claims to prefer Spotify streams upsampled to DSD through the Lumin over high res streams. That is quite a big claim. If I also found that I preferred Spotify streams upsampled to DSD over streams from Qobuz/Tidal, I would pretty quickly drop my Qobuz, Tidal, and Roon subscriptions. But based on my experience with upsampled DSD via HQ Player, I don’t expect to prefer upsampled DSD via Lumin to native high res PCM.

One takeaway from Huff’s video, however, was that an upsampler placed at the streamer level or farther downstream, would benefit all lossy content -- which arguably would benefit form upsampling the most. My wife and I both love Spotify and haven’t made the jump to Apple Music even though we have Apple devices. Although Apple Music contains high res content, I don’t like Apple Music’s policy of deleting my saved music, artists, and playlists if I ever take a hiatus from it’s subscription.

Regarding HQ Player, I found HQ Player to be a laggy and a clunky experience generally. I wasn’t a fan of waiting 5 seconds for music to begin. Sometimes I would press play and nothing would happen. I used HQ Player to see if DSD was something I would get into, but I preferred the sound of PCM to (upsampled) DSD on my system. And, when I used HQ Player to upsample CD quality tracks from Tidal into high res, I ended up preferring those filters that sounded the same as CD quality streamed from Tidal through Roon.

I never did get around to installing HQ Player onto a computer to sit between the Roon Core and streamer, which may solve the inconsistent experience or reduced lag. I also haven’t installed HQ Player on an M-chip Apple device to see if that removed lag as someone claimed. HQ Player allowed me to experiment with my system and get a taste for (upsampled) DSD. I also found it fascinating how different filter/dither options in HQ Player could alter the sound.

As described above, I understand that both the PS Audio DirectStream Mk2 DAC and HQ Player up sample PCM to DSD. Obviously, some people believe this conversion results in improved sound quality and others do not. I have no DSD listening experience so I do not know. I also understand that the other components in the system, obviously, influence the overall sound quality (as described above). My questions are:

1) Are there other devices that also up sample PCM to DSD? For example, can I purchase a device that accepts PCM out from my Aurender, for example, converts it to DSD for input to my DAC that supports true 1-bit DSD for DSD content? Does this really MAKE sense?

2) How exactly does PCM get converted to DSD? I know the PS Audio DirectStream DAC uses FGPA with custom instructions to make this conversion. I assume a combination of software and hardware are required but, in simple terms, how exactly does this conversion work? How much computing power is required for a PCM to DSDS conversion?

3) Once the PCM stream has been converted to DSD, how does the DSD stream get converted to analog? Does it use a filter to make this conversion? Please explain?

4) The most important question is how does listening to PCM streams compare to listening to a DSD Stream that has been converted by one of the above devices?

5) My current audio system only streams PCM so I am looking for more information on these PCM to DSD conversion options. Thanks.

 

Not DSD or PCM specific, but I've been finding more and more that it's dependent on where the recording comes from which is hard if not impossible to determine via streaming. Some 16bit recordings can sound phenomenal and in some cases, better than 24bit+ versions of the same song/album.

 

This isn't as much an issue if you use physical or local media, of course.

While not crazy expensive, I do have a decent system (MSRP > $100k) with a tube preamp and SS monoblocks.

I find HQP to be the best value for money purchase in my system. BTW: HQP5 is a big upgrade from HQP4 in terms of SQ.

While I would say, in high end (and / or resolving) systems, the benefit is system dependent, a system change can change preference from not using HQP to using it (or vice versa).

Here’s a specific example: recently, with a VAC preamp and Kubala Sosna Fascination speaker cables, I preferred Roon redbook (no HQP). When I changed to KS Elation cables, I prefer using HQP over redbook (Elation, while an upgrade over Fascination, makes my system sound a tad less musical than before. Upsampling to DSD512 using AMSDM7EC 512+fs / Sinc mGA brought back the musicality in spades.

 

 

I cannot and shall not have any general conclusion.  I shall say though that HQ Player upscaling all PCM to DSD only using HQ Player author Miska's specific recommended modulator (requires high power dedicated computer) > T+A DAC 200 ($7150 USD SRP) in DSD NOS (non-oversampling) mode rivals any cost no object digital playback system.  I would not be afraid to compare it to a $100k MSB or dCS rack (I heard an $80k MSB rack and the latter.)  

Regardless all the different potential perceptions and preferences experienced listeners may have for DAC A vs. DAC B: I shall say here and now that every single experienced listener employing a decent system or better who AB tested (takes only 20S to switch) PCM (even HR) vs. DSD on this system would prefer the latter every single time and without exception on all music programs.  DSD's advantage in this system is well beyond the pale correlating to different taste.  DSD appears to have 6-10 dB greater dynamic range just to start and no one in a quiet room prefers dynamic compression over the lack of same.  

The primary difference in the above test giving DSD the immeasurable edge every time over any PCM is the fact that all the filtering, modulation and DSP takes place in DSD octaves farther from the audible spectrum than with PCM.  

All bets off if HQ Player is not setup appropriately for DAC 200 or if a different DAC is employed.

Note I love John Darko's advice in general and am NOT contradicting his general advice.  My advice is very specific first hand.  If anyone wants to hear the above described AB test I'm 75 minutes NE of Salt Lake City.  The rest of the system is world class including the room and cables.  

I agree with Paul McGowan that DSD is superior to PCM. Part of the reason is that when DSD hits a DAC it doesn't go through filters like PCM does. One of the reasons why a 24/192 recording sounds even better when transferred to DSD to make an SACD. Chandos even has a recording of Symphonie Fantastique that was recorded at 32/352.8 and then transferred to DSD64 for the SACD. And there's another way to listen to DSD off SACDs - take the HDMI out of a universal player (Oppo, Sony. Reavon...) to a GeerFab Audio D.BOB, which legally extracts the DSD and hi-res PCM and outputs over coax to a DAC that accepts DoP over coax (Mytek, Benchmark, Chord, Topping...). (Disclaimer - I'm the inventor of the D.BOB.) Upsampling - I use an iFi Audio Pro iDSD to upsample Internet radio and CDs to DSD1024. Breathed much needed new life into my CD collection. Not quite SACD quality, but a whole lot better than CD.

Of course filters influence the sound of oversampling DACs. But filter algorithms in all DACs are all built around common standards with a finite list of variables they can play around with, and there’s nothing particularly innovative or earth-shattering between one DAC manufacturer’s vs another’s. In the case of brands that share the same DAC chipset (such as an ES9039pro), they’re all defined by the options hard coded into the chip, and are just given different names.

Regardless, there’s certainly nothing about PS Audio’s filters that are any different from those offered in other DACs, and there’s definitely nothing magical about them that warrants a $6,000 price tag. Anyway, the discussion of filter options is not germane to my point about the myths of IIS being an allegedly superior digital audio transport or about the outrageous price points given to these “boutique” DACs that don’t measure any differently than a cheap delta sigma SMSL 400ES or a Topping d90se. In fact the affordable DS DACs of the past few years measure demonstrably better than the offerings from the likes of PS Audio and Chord, and they absolutely trounce the measurements of archaic, pricey R2R ladder options.

This phenomenon isn’t something that leads to any subjective differences in sound quality that would convince the acolytes away from the high dollar boutique brands, because we reached the limits of the human ear years ago with digital to analog conversion. An eight year old Chord DAVE sounds every bit as good as a Topping d90se because the Topping’s superior distortion scores and noise floors are not audible (to this point, even though I somewhat hypocritically demand DACs capable of PCM 768 and DSD 512 personally, anything above a SINAD of 96 or a shade above 16 bit is completely inaudible, and the human ear has been proven to be incapable of discerning resolutions much higher than Red Book CDs).

So in spite of the fact that we’ve scientifically proven that advances in digital technology have now allowed mass production of DACs that provide the highest possible audio reproduction that can be achieved, outfits like PS Audio are going to keep writing ever more inscrutable and technologically advanced-sounding one sheets that will keep convincing their customers to shell out wads of cash for the “latest” in DAC technology, despite the fact that the technology reached its finite limits in the last decade and there’s no place left to go.

The future of the audiophile industry, when it comes to DACs, is wholly dependent on cognitive bias and the reliable consumer base that is psychologically held captive to the myth that high cost correlates with quality. Given that DACs are nothing more than computers, why else would otherwise sensible folks still shell out $14,000 for a Chord DAVE in 2023, when the the device’s tech is almost nine years old? Think about other digital devices with nine year old tech—2014’s cell phones, computers, televisions, etc, and ask yourself whether you’d pay ultra premium prices for one of THOSE. 2023 prices for the iPhone 6–take a moment and just think about the motives of ritzy audio manufacturers vs common sense, for the sake of your wallet if nothing else!

As a friendly aside, factors such as soundstage, presence and imaging (I have to set aside “realism” because I think that term is undefinable) are all primarily a matter of the headphones used to play back audio streams. There is no way for any of these factors to be packaged into a digital audio stream, and again, if you think logically about the function and purpose of the digital to analog conversion process, there is no way these attributes could exist at this stage of audio processing. A DAC’s purpose is to recreate in an analog format what existed in the original master when it was sampled into a digital format. If you can explain to me how “soundstage” could be carried forward in this format, I’m anxious to hear it. 
 

There’s a great deal of misappropriating analog terms into the digital realm in these rooms. These nebulous claims are exactly what manufacturers want us to do to convince ourselves that they actually mean something when they claim that their DAC “takes our listening experience to another level”. I think it’s at great risk to our wallets that we fall for these unsubstantiated concepts that defy the logic of how digital audio works. There is no “sound” in a digital audio stream until it reaches our amplifiers—there’s only an encoding of the original master, full stop.

I wasn’t being snarky in my post, so I’d appreciate not being condescended to or demeaned in turn, if I may so ask. I’m well aware of PS Audio’s coup in circumventing Sony’s DRM licensing limitations. Perhaps a little explanation of the back story of this stroke of genius is in order so folks won’t continue to mistake it for “innovation”. 

Sony is the original inventor of SACD technology, as you know. And since the inception of the format, Sony has forced all manufacturers of SACD players to follow their own hardware limitations, namely downsampling all digital audio streams to a maximum of 48 kHz and 16 bit. Sony also owned a majority of the masters originally released in the SACD format, and they invoked this restriction to prevent bootleggers from pirating their own digital masters that far exceeded the resolution of standard Red Book CDs. 

You’ll note that PS Audio, the only outfit who has managed to circumvent these restrictions, did not release their DirectStream DAC in isolation—they did so with the simultaneous release of first their PerfectWave SACD player, that was designed to offer this benefit when coupled with one of their proprietary DACs, via McGowan’s beloved IIS protocol. Then they “upgraded” the technology to their DirectStream line, at even higher prices. 

PS Audio could not have achieved this “innovation” without the express permission and cooperation from Sony, without incurring major lawsuits. Even high dollar Marantz SACD players do not have access to the bit perfect DSD layer, because they are just as restricted by Sony’s DRM rights as any other manufacturer is. So PS Audio did not achieve this breakthrough by any strokes of technological brilliance—they merely bought the rights to supersede the licensing limitations from Sony, who would only agree to this if the format was limited to a non-standard (at the time) transmission protocol—namely, IIS. 

So that $6,000 price tag you paid for to gain access to that DSD layer largely went  to Sony for their licensing rights. There’s no magic otherwise in PS Audio’s DAC or SACD player, nor is there anything superior about IIS over standard USB (despite McGowan’s inscrutable claims to the contrary). It’s all a ruse to get you believe that McGowan has mystical gifts as an engineer, and to convince you to part ways with a month’s worth of salary to buy these two units as a pair accordingly.

Lest you misunderstand my motives, I’ve long been tempted to buy into this just so I could access the only way possible to play my SACDs natively through my headphone setup. If you’re happy with these assets, as I’ve said before, I champion and celebrate that for you, and I have aspired to what you have myself for some time. Where I take exception is when folks start implying that the DAC itself offers superior performance over much less expensive offerings, or when folks pass along McGowan’s unsubstantiated scientific claims that he promotes to obfuscate the truth behind DirectStream’s technology, the “benefits” of IIS, and his profit-seeking motives. PS Audio is inarguably savvy, but their ingenuity is in strategy, not so much on the bench. Peace.

@srkbear Thanks for trying to save us all from the snakey oil salesmen out there. It's highly appreciated.

However Paul McGowan's SACD/CD transport is a ground breaking product that outputs pure native DSD over I2s to a DAC, eliminating and freeing the user from restrictive conversion processes.

As far as the sound of DACs is concerned it's more of how much more detail, soundstage, presence, realism, imaging etc., not just warm or cold.

We don't tailor sound using amplifiers and other components, we look to find a synergy that will bring us as close to the live event as possible. Whether that's a concert or studio session.

I’m relatively new to this forum, and I’m certainly not looking to make any enemies here. I’ve always maintained the view that if someone is happy with the sound of their audio system, and it gives them joy listening to music, that is a beautiful thing—and it would be a self-serving, destructive act to attempt to sully that experience for them by criticizing their choices.
 

However, I see a lot of talk on here about the “sound” of a DAC, particularly in favor of PS Audio’s $6,000 take on this technology, and Paul McGowan’s “proprietary” terms he invokes to describe its ephemeral results. And it is a very different thing when one is suggesting to others that the indefensibly priced DirectStream is “the only way to go” if you want to achieve the pinnacle of sound, from a component that shouldn’t have a “sound” to begin with—especially when their claims are as scientifically baseless as the language used by the marketing mouthpiece of the manufacturer himself. Has anyone ever asked themselves what factors justify that price point, for a device that exists almost exclusively in the digital realm?
 

What exactly should a DAC “sound like”? A DAC is nothing more than a computer, designed to convert a digital sample of an original master back into its original form with the least amount of errors possible—errors that take the form of noise, distortion or “jitter” when our amplifiers transmit the signal to our headphones or speakers. If a DAC is doing its job properly, it shouldn’t have a “sound” other than what was intended in the mastering process. If it sounds “warmer” than other DACs, assuming that all other components are unchanged, then that’s either a product of distortion (errors) or of the cognitive bias introduced by having spent $6,000 on the device itself. Logically (and therefore scientifically), the DAC should be an invisible part of the signal chain—reproducing the original master as faithfully as possible, allowing us to tailor the sound to taste in our choice of amplifiers, DSPs/ASPs and headphones.  
 

For those in the market for a DAC, or an audio system in general, please do your research before spending your hard-earned cash on your choice of components, and please be very wary of the advice you receive from others, particularly when they are suggesting to you that you would be a pinhead if you didn’t buy into the marketing claptrap from “premium” brands. A very wise audio engineer once told me to spend my hard-earned cash in the following order of cost, from least to most—and it has always served me well: Cables—>Streamer—>DAC—>Amplifer—>Headphones. The latter component is the most personal one, the one that has the most significant impact on the resulting sound. Find a DAC that has the best measurements (lowest noise, distortion and error rate)—if you do your homework you’ll find that recent advances in technology have brought us options that require a minimum of expense to do the job exceptionally well. Peace.

This (can be) a pretty involved topic in so far as understanding the two different formats and especially upsampling in the correct context, since it (upsampling to DSD) is actually fixing / addressing inherent distortion

"Most" DAC’s that use chips (ESS Sabre for ex) as well as CHORD, upsample everything internally and almost impossible to avoid (there are some exceptions and depending on how the DAC designer implements the "chip" can disable certain functions). So the comment above about "cheaper equipment because the upsampling algorithm is basically filling in holes" and "Therefore the upsampled music is not true to source" is incorrect IMHO.

NOS DAC’s (Holo, Mustec, T+A, last gen Lampi’s) are just that and there are a fair amount of NOS DAC’s available nowadays. These are a prime target for umsampling prior to DAC and a widely used method at this time is HQP. Its a great software with many filters and modulators for both PCM + DSD. The developer is all about DSD (upsampling PCM > DSD and native DSD is HQP wheelhouse) and "usually" best if installed on its own dedicated wkst. If using roon (for lib mgmt and sending to HQP), keeping roon on its own dedicated core wkst is a good idea.

The format hi-res is really irrelevant and is used incorrectly in a lot of topics IMHO. It all has to do with provenance > the masters. There are a lot of great redbook sounding CD’s as well as a lot of crappy sounding re-masterd and upsampled and sold as hi-res from same CD.

If the master / recording was done natively in a higher rate (24/96 for example) or DSD, that’s great and more then likely it will sound very good (or as best as the mastering). My point is the bit rate isn’t automatically the end all - you have to know where it came from (to be absolutely certain). I would venture to say many artists now record in 24/96 and very few in native DSD.

DSD IMHO is anything but hard, brittle etc.. and generally speaking my fav format especially on a DAC meant for it.

The use of the new HQP EC modulators and upsampling to DSD 256 is pretty ridiculously good and I upsample everything to DSD 256 or 512 with the orig modulator.

This is all really to say, there is no right or wrong way to go here and comes down to personal preference and ones system.

If you have a solid wkst you can run HQP on with a quality USB card to go DAC direct OR send HQP to an NAA (opticalRendu, Pi, NUC, Fillet etc..) your golden. Additionally, you can now very easily get HQP in an embedded version (custom ubuntu with HQP installed) OS image. Just burn it to your internal M.2 or boot to a USB stick.

 

A few takeaways from the discussion so far:

  1. People here have a strong preference for a format, whether it’s PCM or DSD. That tells me DSD is something I need to hear, and I can decide whether I prefer it or not. My current DAC doesn’t support DSD, but all of the upgrades I’m considering do, so I will trial DSD at some point in the future.
  2. People with PS Audio’s DAC-based upsampling solution seem to greatly enjoy it. However, server-side upsampling doesn’t seem to be as polarized--for or against. I’m in the process of replacing the rug in the listening room. Hardwood is exposed for now, but I may try A-B’ing Roon’s (PCM) upsampling before and after with the hardwood, and later with the rug.
  3. I just ordered the components to convert cable to fiber right before internet enters the streamer. I’ll try A-B’ing Roon’s (PCM) upsampling before and after inserting fiber as well. If fiber has the effect I hope it does, fiber may help further reveal the benefits/detriments to upsampling.

I’m expecting fiber to make a big positive difference. But from this conversation, I’m not certain whether Roon’s upsampling will be a net improvement. I have a lot of variables to control for, so I’ll be a bit longer before reporting my findings.

I've been using a PS Audio DirectStream Sr. for a few years now and although I've auditioned quite a few other units, none gets close. The DSS has an organic, almost analog, sound that is truly intoxicating. Feed it a native signal (no upsampling) and it's glorious. If you upsample via Roon (which what I have and use), you can clearly hear that it adds a digital "edge" to the sound. Sounds more detailed at first, but further listening and you realize it's just a harder leading edge.

 

If you really want to hear the best possible DAC, get a DirectStream Sr, upgrade the output transformers to Edcor XS4400 and get an external power supply for the digital side, letting the DSS only power the analog side. That beats everything out there for a relatively cheap investment.!!!

If you have a quality DAC and equipment you shouldn't feel the need to upsample.

Upsampling is OK I guess when you have cheaper equipment because the upsampling algorithm is basically filling in holes with guesses produced by the algorithm.

Therefore the upsampled music is not true to source.

In my many years of experience with DACs and digital formats DSD is indeed best. And PS Audio's Direct Stream is the best sounding DSD DAC I've heard. Almost always a noticeable improvement of PCM files using it compared to directly through a high quality PCM DAC (both with and without the fiber conversion trick, which I also recommend).

I also highly recommend their SACD transport using I2C to send the DSD layer directly to their DAC.

Don't  care for converting pcm to dsd with my setup and Roon. Staying with pcm for pcm recordings most revealing and tranparent, dsd conversion sounds more like hifi, smooths things over, micro dynamics lost, sense of live performers in room diminished. YMMV.

 

And yes, do the fiber conversion, you may find you don't need or care for extra processing after this.

 

My take is processing makes noise, the more processing the more noise. Many off the shelf servers don't have the computing power to run high level processing, one should ask themselves why. Is it because they've determined costs are greater than benefits?

FI know in the past years casting Qobuz that I notice the difference between 16-bit and 24-bit content. But I haven’t picked up on SQ improvements based solely on sample rate.

Yup. I notice a notable difference between 16/44.1 and 24/96 — not so much between 24/96 and 24/192. Not a fan of SACD. Something just sounds “over processed” with a resulting thin veil compared to PCM and hi-res PCM that sound more transparent. But, I’ll also say that well-recorded 16/44.1 stuff sounds just as good as some 24/96 stuff, so go figure. For me, I’d NEVER convert PCM to DSD — it just seems like more unnecessary processing and original signal degradation.

Anyway, there are many new DACs out today that are multi-bit R2R and non-oversampling (NOS) precisely because it avoids all the extra processing of 1bit and upsampling/oversampling DACs. Many people find these multi-bit DACs to sound more natural and analog sounding and less “digital” sounding. I’m personally agnostic and think either technology can sound great, but I recently acquired an R2R DAC and absolutely love it. In short, while upscaling in video yields pretty universally praised benefits, in 2-channel audio upsampling/oversampling and the brick-wall filters involved with single-bit DACs versus NOS multi-bit DACs is a bit more controversial. Single-bit devotees will tell you they measure better, and multi-bit fans will tell you they sound better — not dissimilar to the ongoing tube vs. solid state debate. Pick your poison, but I just wanted to point out the relatively recent popularity of R2R and NOS DACs (some R2R are NOS and some are not) that fly in the face of the upsampling/oversampling tech. To me, there’s a reason there are more and more R2R/NOS DACs hitting the market. Sorry to maybe complicate this for you, but hope you find this helpful and best of luck in your search.

When I get a chance to listen critically, I'll post my findings here. I know in the past years casting Qobuz that I notice the difference between 16-bit and 24-bit content. But I haven't picked up on SQ improvements based solely on sample rate. This might be because bit depth variable has always been easier to test for. This will be my first time intentionally isolating sample rate. 

This leads me to another question, how might I go about upscaling the bit depth? (Is "upscaling" the right word here?) I can't easily attach screen shots here to show, but Roon's "sample rate conversion" area only seems to affect the sample rate, not the bit depth. 

I looked into upsampling with Roon. When I use the computer as Roon Core only (and send the stream to the HiFi), I don’t see the ability to upsample. I do see the ability to upsample when the computer is also acting as the streamer.

Right click on the device at the lower right, select Device Setup, then scroll down and click on Advanced Settings.  Set the maximum sample rate there. 

Then right click on your device again but select DSP.  Second slider will be Sample Rate Conversion.

 

 

I love DSD.  My Oppo 105 outputs the DSD layer of SACD over HDMI into the HDMI input of my Bryston DAC3.  I also have some DSD downloads but most of my DSD listening is SACD.  

Roon Roc Nuc upsampling to 256 DSD feeding a VMV D2 is not ‘bright or hard’, but smooth and quiet with loads of detail and great separation and layering. Backup vocal singers are beautiful and easy to follow on say a John Gorka selection. The AKM 4499 shines in this configuration. Everything has a palpability, certainly no ‘edge’ running native. Plenty to fiddle about with without going into HQPlayer. PCM can also be very good and once you find a preference it is easy to swap in or out to various resolution from your remote device. Build a Nuc, control what you plug it into, how you feed it, (digital likes isolation, ssd, nice filter caps, ect…) mine does DSD512 also, but consensus seems 256 is the sweet spot. I2S and better clocking thru a DDC, I’m told offer further improvements. Streaming radio, something like ‘Mountain Chill’ out of Telluride, from Roon sounds damn good, as it looks to be upsampled from its lower decent resolution broadcast, then forwarded to the dac. Surprisingly nice sound quality, upsampled from around the world, via live stream. That’s a pretty big plus for Roon…

http://mountainchill.com/

@deadhead1000 I've been trying to avoid introducing general purpose computers into the audio chain, but when Spotify ditched it's CD quality tier I caved and picked up Roon and Qobuz. I still don't particularly like having to boot up a computer in a different room and launch Roon on it to be able to cast Qobuz to the HiFi--not as convenient as I'd like. But I recently read about someone converting their internet to optical to isolate their server from the streamer. It's a $100 project. I'm going to try that out too. 

DirectStream is the only way to go. I guess you missed the 20% off plus big trade INS a few years back. I traded in a buncha junk and got it new for about $3.3K.

Stay far away from PCs.

@erik_squires I looked into upsampling with Roon. When I use the computer as Roon Core only (and send the stream to the HiFi), I don't see the ability to upsample. I do see the ability to upsample when the computer is also acting as the streamer. I've searched the Roon forums, and that's all I've seen. If you know of a setting I've missed, please share. 

@jgoldrick I am interested to try upsampling PCM to DSD. Looks like a bit of a learning curve to use HQP generally, but I'm hoping that once the settings are all dialed in that I won't have to fiddle with the software much after that. We'll see how it goes. 

It's funny to me that USB cables become relevant again to attain DSD. As my system is graduating from entry level gear, this next tiers of DACs mostly have DSD decoding abilities. 

 

I have used HQPlayer for about 2 years.  I really like it.  I upsample to 192K PCM and use the DSP FIR section (REW and rePhase - kindof a noob at that).  Anyway, if you are planning to upsample to DSD, you will need some power, depending on how high up the DSD ladder you go.  At least 4 core I5 for a CPU, I suspect, and a good chance of a $1500 or more video card.  HQPlayer can use the GPU for various functions.  I suggest you research the requirements for a computer/device to run HQPlayer and the desired DSD rate before you buy.

I was researching building a better PC than I have now just for HQPlayer and it fairly quickly hit $3-$5K fast, and that was not a high powered machine.  That's when I decided I was going to stick with PCM and re-aligned my purchases accordingly.

Roon has built-in upsampling, so that’s one option. Here’s what I think:

About 15-20 years ago there was a revolution in how well your average DAC chip handled 44.1kHz/16 bit audio, and it’s now very very good.  The big delta between high resolution music and CD's has all but vanished.

While it is impossible to add new data to audio signals for now, barring some future AI process, higher sample rates do behave differently in the very top octave.

I don’t like DSD for the most part. I find it bright and hard.

 

I went with the PS Audio Direct Stream, moving to it for the upsampling and FPGA. As a techie, I liked the idea of FPGA and to me upsampling makes sense. I looked at the Chord, but yeah, way more money and extra cables.

I don't have any experience with HQ player, but based on your $5K system (nothing to sneeze at, it took me 20+ years to get to that level!) a $6K DAC is hard to justify. However, personally I stayed away from the PC and software route. True, plenty of people love that type of system, but to me it introduces too many variables and places to inject noise.

Last CD vs. Hi-Res? Depends on the level of your system I was fine with MP3's in the 90's (GASP!!)....but now have only CD level and slowly moving my favorite stuff to Hi-Res. 

Hope this helps a bit.